The objective of the present study was to assess the relative efficacy of insulin or glibenclamide treatment for non-insulin-dependent diabetes mellitus (NIDDM) over 42 months. We performed a randomised, controlled trial allocating patients treated with diet and oral antihyperglycaemic agents to treatment with glibenclamide or insulin to achieve HbAlc levels under 7.5%. We included 36 subjects with established NIDDM of more than 2 years' duration. Mean HbAlc levels were significantly reduced in patients allocated to insulin treatment from 9.1 +/- 1.4% before the start to 7.8 +/- 1.3% (p < 0.05) after 1 year, and did not change significantly thereafter throughout the study period. Mean HbAlc levels increased during the study in the patients allocated to glibenclamide treatment, and 11 of 18 patients had to be switched to insulin treatment due to increasing hyperglycaemia (HbAlc > 10%). Mean body weight increased in the subjects allocated to insulin by 7.2 +/- 4.1 kg during the study period. In conclusion, insulin was more effective than glibenclamide treatment in obtaining control over hyperglycaemia in these patients, and once improved, glycaemic control did not deteriorate over 42 months in the insulin-treated group. Two thirds of the patients allocated to glibenclamide treatment had to be given insulin due to inadequate glycaemic control.